handler.py

import torch
from typing import Dict, List, Any
from transformers import (
    AutomaticSpeechRecognitionPipeline,
    WhisperForConditionalGeneration,
    WhisperTokenizer,
    WhisperProcessor,
    pipeline
)
from peft import LoraConfig, PeftModel, LoraModel, LoraConfig, get_peft_model, PeftConfig

class EndpointHandler():
    def __init__(self, path=""):
        # Preload all the elements you are going to need at inference.
        # peft_model_id = "cathyi/tw-tw-openai-whisper-large-v2-Lora-epoch5-total5epoch"
        peft_model_id = path
        language = "Chinese"
        task = "transcribe" 
        peft_config = PeftConfig.from_pretrained(peft_model_id)
        model= WhisperForConditionalGeneration.from_pretrained(
            peft_config.base_model_name_or_path
        )
        model = PeftModel.from_pretrained(model, peft_model_id)
        tokenizer = WhisperTokenizer.from_pretrained(peft_config.base_model_name_or_path, language=language, task=task)
        processor = WhisperProcessor.from_pretrained(peft_config.base_model_name_or_path, language=language, task=task)
        feature_extractor = processor.feature_extractor
        self.forced_decoder_ids = processor.get_decoder_prompt_ids(language=language, task=task)
        self.pipeline = AutomaticSpeechRecognitionPipeline(model=model, tokenizer=tokenizer, feature_extractor=feature_extractor)
        # self.pipeline = pipeline(task= "automatic-speech-recognition", model=model, tokenizer=tokenizer, feature_extractor=feature_extractor)
        self.pipeline.model.config.forced_decoder_ids = self.pipeline.tokenizer.get_decoder_prompt_ids(language="Chinese", task="transcribe")
        self.pipeline.model.generation_config.forced_decoder_ids = self.pipeline.model.config.forced_decoder_ids  # just to be sure!

    def __call__(self, data: Dict[str, Any]) -> List[Dict[str, Any]]:
        """
       data args:
            inputs (:obj: `str` | `PIL.Image` | `np.array`)
            kwargs
      Return:
            A :obj:`list` | `dict`: will be serialized and returned
        """

        inputs = data.pop("inputs", data)
        with torch.cuda.amp.autocast():
            prediction = self.pipeline(inputs, generate_kwargs={"forced_decoder_ids": self.forced_decoder_ids}, max_new_tokens=255)
            # prediction = self.pipeline(inputs, return_timestamps=False)
        return prediction